arXiv subjects

Mariana Fernandez-Espinosa

Publications and source records attributed to Mariana Fernandez-Espinosa.

5 recordsLinked to original sources

arXiv2026

ViT-Explainer: An Interactive Walkthrough of the Vision Transformer Pipeline

Transformer-based architectures have become the shared backbone of natural language processing and computer vision. However, understanding how these models operate remains challenging, particularly in vision settings, where images are processed as sequences of patch tokens. Existing interpretability tools often focus on isolated components or expert-oriented analysis, leaving a gap in guided, end-to-end understanding of the full inference pipeline. To bridge this gap, we present ViT-Explainer, a web-based interactive system that provides an integrated visualization of Vision Transformer inference, from patch tokenization to final classification. The system combines animated walkthroughs, patch-level attention overlays, and a vision-adapted Logit Lens within both guided and free exploration modes. A user study with six participants suggests that ViT-Explainer is easy to learn and use, helping users interpret and understand Vision Transformer behavior.

cs.CV↗

arXiv2025

Practicing a Second Language Without Fear: Mixed Reality Agents for Interactive Group Conversation

Developing speaking proficiency in a second language can be cognitively demanding and emotionally taxing, often triggering fear of making mistakes or being excluded from larger groups. While current learning tools show promise for speaking practice, most focus on dyadic, scripted scenarios, limiting opportunities for dynamic group interactions. To address this gap, we present ConversAR, a Mixed Reality system that leverages Generative AI and XR to support situated and personalized group conversations. It integrates embodied AI agents, scene recognition, and generative 3D props anchored to real-world surroundings. Based on a formative study with experts in language acquisition, we developed and tested this system with a user study with 21 second-language learners. Results indicate that the system enhanced learner engagement, increased willingness to communicate, and offered a safe space for speaking. We discuss the implications for integrating Generative AI and XR into the design of future language learning applications.

cs.HC↗

arXiv2025

When Technologies Are Not Enough: Understanding How Domestic Workers Employ (and Avoid) Online Technologies in Their Work Practices

Although domestic work is often viewed as manual labor, it involves significant interaction with online technologies. However, the detailed exploration of how domestic workers use these technologies remains limited. This study examines the impact of online technologies on domestic workers' work practices, perceptions, and relationships with customers and employers. We interviewed 30 domestic workers residing in the United States, who provided examples that highlight the insufficient transformative role of current online technologies in their work. By conducting a thematic analysis, we characterize how they approach and avoid these digital tools at different stages of their work. Through these findings, we investigate the limitations of technology and identify challenges and opportunities that could inform the design of more suitable tools to improve the conditions of this marginalized group.

cs.HC↗

arXiv2025

Augmenting Teamwork through AI Agents as Spatial Collaborators

As Augmented Reality (AR) and Artificial Intelligence (AI) continue to converge, new opportunities emerge for AI agents to actively support human collaboration in immersive environments. While prior research has primarily focused on dyadic human-AI interactions, less attention has been given to Human-AI Teams (HATs) in AR, where AI acts as an adaptive teammate rather than a static tool. This position paper takes the perspective of team dynamics and work organization to propose that AI agents in AR should not only interact with individuals but also recognize and respond to team-level needs in real time. We argue that spatially aware AI agents should dynamically generate the resources necessary for effective collaboration, such as virtual blackboards for brainstorming, mental map models for shared understanding, and memory recall of spatial configurations to enhance knowledge retention and task coordination. This approach moves beyond predefined AI assistance toward context-driven AI interventions that optimize team performance and decision-making.

cs.HC↗

arXiv2025

Breaking the Familiarity Bias: Employing Virtual Reality Environments to Enhance Team Formation and Inclusion

Team closeness provides the foundations of trust and communication, contributing to teams' success and viability. However, newcomers often struggle to be included in a team since incumbents tend to interact more with other existing members. Previous research suggests that online communication technologies can help team inclusion by mitigating members' perceived differences. In this study, we test how virtual reality (VR) can promote team closeness when forming teams. We conducted a between-subject experiment with teams working in-person and VR, where two members interacted first, and then a third member was added later to conduct a hidden-profile task. Participants evaluated how close they felt with their teammates after the task was completed. Our results show that VR newcomers felt closer to the incumbents than in-person newcomers. However, incumbents' closeness to newcomers did not vary across conditions. We discuss the implications of these findings and offer suggestions for how VR can promote inclusion.

cs.HC↗